Daniel Mendelevitch

research @runwayml

[Daniel Mendelevitch @danielmend_]

@MiSunKw @runwayml the fun fact is that i have no fun facts

[Daniel Mendelevitch @danielmend_]

life update: i joined @runwayml this week :DD

[Luma @LumaLabsAI]

Today, we release Inductive Moment Matching (IMM): a new pre-training paradigm breaking the algorithmic ceiling of diffusion models. Higher sample quality. 10x more efficient. Single-stage, single network, stable training. Read more: https://t.co/pewpFNtJAX

[Luma @LumaLabsAI]

Image to Video with #Ray2 is here. Drop any image into Dream Machine to make it a video and bring it to life — from historic artifacts and paintings to memes, custom art and photos of any kind. Available now.

[Daniel Mendelevitch @danielmend_]

check out our new video model😄!!!

[Luma @LumaLabsAI]

Introducing Ray2, a new frontier in video generative models. Scaled to 10x compute, #Ray2 creates realistic videos with natural and coherent motion, unlocking new freedoms of creative expression and visual storytelling. Available now. Learn more https://t.co/jGI6KmRQpR.

[David McAllister @davidrmcall]

Decentralized Diffusion Models power stronger models trained on more accessible infrastructure.

DDMs mitigate the networking bottleneck that locks training into expensive and power-hungry centralized clusters. They scale gracefully to billions of parameters and generate photorealistic images with just a week of training on eight independent GPU nodes. They’re easy to implement, adopt DiT hyperparameters directly and outperform standard models FLOP-for-FLOP.

[Daniel Mendelevitch @danielmend_]

👀

[Luma @LumaLabsAI]

Are you ready?

[Luma @LumaLabsAI]

Say hello to the all-new #DreamMachine. 🚀 Your visual thought partner—where ideas become reality. Ideate, visualize, and share your ideas with the world. Available now.

[Daniel Mendelevitch @danielmend_]

life update: i started interning at @LumaLabsAI today :D

[Maciej Kilian @kilian_maciej]

Excited to announce our latest paper analyzing the computational tradeoffs between image synthesis approaches. We compare the text-to-image capabilities of diffusion, next-token, and masked-token prediction across various compute budgets. We also briefly investigate the impact of several modules in the recipe - the autoencoder, conditioning method, and usage of EMA on the model weights.
https://t.co/LMihQsHu8f

[Daniel Mendelevitch @danielmend_]

@daniel_c0deb0t @AnthropicAI congrats!!

[Daniel Mendelevitch @danielmend_]

@_neelr_ @_pranavnt @lnofx seqlen/ctx len is 77, the embeddings themselves are higher dimensional

[Daniel Mendelevitch @danielmend_]

holy fuck

[Tim Brooks @_tim_brooks]

Sora is our first video generation model - it can create HD videos up to 1 min long. AGI will be able to simulate the physical world, and Sora is a key step in that direction. thrilled to have worked on this with @billpeeb at @openai for the past year https://t.co/p4kAkRR0i0

[Cristóbal Valenzuela @c_valenzuelab]

game on

[Daniel Mendelevitch @danielmend_]

@dluozhang @samueloshay nobody has ever said this, like ever

[Daniel Mendelevitch @danielmend_]

@dluozhang the el segundo canes should be a holy site

[Daniel Mendelevitch @danielmend_]

@TheStithLord dude this is so cool

[Daniel Mendelevitch @danielmend_]

@DigThatData @CoreWeave @StabilityAI congrats!

[Daniel Mendelevitch @danielmend_]

@unixpickle 🤷‍♂️

[Sumith Kulal @sumith1896]

Real-time high-quality sampling is a totally different experience, try it out here: https://t.co/w5VaCITg4B